What problem does it solve? When a long AI coding session ends or compacts, the reasoning and context die with the conversation, forcing the next session to re-derive decisions or follow stale leads. This Skill packages the current project state into a dense, verified prompt the next session can paste to resume work with zero re-learning. ## Core Features & Use Cases - Source-of-truth grounding: Reads think-and-ship state (roadmap_status, roadmap_next, ship_status, signal_pending) plus the current code to determine what is actually true now, not what the trace once claimed. - Ruthless trace filtering: Mines the think trace for the 'why' and hard-won gotchas while excluding superseded decisions, abandoned branches, and stale next_actions. - Fact verification: Confirms every load-bearing fact (files, recipes, gate commands) against the current code before carrying it forward, labeling anything unverifiable. - Use Case: At the end of a long refactoring session, run /handoff to produce a single fenced prompt naming the exact next roadmap chunk, proven templates to clone, verified gotchas, real gate commands, and honest open items for the next session. ## Quick Start Ask the AI to write a handoff for the next session using the current think-and-ship roadmap and ship state.
